South Africa District HIV Estimates

South Africa District HIV Estimates combine district-level data about HIV from several sources in a statistical model:

Prevalence (15-49) & PLHIV (15+)

ART coverage & Number on ART (15+)

Incidence (15-49) & New infections (15+)

About

Estimates are developed through a technical collaboration of:

the MRC Centre for Global Infectious Disease Analaysis at Imperial College London,
the Centre for Infectious Disease Epidemiology and Research at the University of Cape Town, and
the Health Economics and Epidemiology Research Office (HE2RO) at the University of Witwatersrand

in partnership with the South Africa Department of Health and UNAIDS. <ALSO CDC? USAID? HSRC? OTHERS?>

Estimates are reviewed by the <INSERT OFFICIAL NAME OF ESTIMATES TWG>.

Full results

Full estimates in spreadsheet format can be downloaded here.

Within the zipped folder, the spreadsheet indicators.csv contains estimates of all indicators stratified by:

Area: national, province, district
Sex: both, female, male
Age group: 5-year age groups, all ages, 0-14, 15-49, 15-64, 15+, 50+, 15-24, 25-34, 35-49, 50-64, 64+, age 0, 1-4
Time: June 2017, March 2020, March 2021

For each estimate, the mean estimate and 95% uncertainty range are reported.

The file boundaries.geojson contains geographic boundaries, which can be used producing maps of estimates.

For further information, questions, or feedback about South Africa District HIV Estimates, please contact:

Dr. Jeffrey Eaton (Imperial College London): jeffrey.eaton@imperial.ac.uk
Dr. Leigh Johnson (University of Cape Town): leigh.johnson@uct.ac.za
Dr. Gesine Meyer-Rath (Boston University): gesine@bu.edu

Relationship to Thembisa model estimates

< INSERT TEXT >

Methods: Naomi model

South Africa District HIV Estimates are developed using the Naomi model. The Naomi model is a small-area estimation model for estimating HIV prevalence and PLHIV, ART coverage, and new HIV infections at district level by sex and five year age group. The model combines district-level data about multiple outcomes from several sources in a Bayesian statistical model to produce consistent estimates of multiple outcomes of interest.

The model focuses on current estimates and short-term one-year ahead projections for HIV programme planning purposes. District-level population estimates by sex and age group are drawn from the Statistics South Africa Mid-Year Population Estimates¹ and adjusted to match the Thembisa province-level population by sex and age group.

The first step is to produce cross-sectional estimates for HIV prevalence, ART coverage, and HIV incidence in June 2017, the mid-point of the most recent nationally-representative household surveys. For HIV prevalence, the model is calibrated to survey data about HIV prevalence by district, sex, and five-year age group from the National HIV Prevalence, Incidence, Behaviour and Communication Survey (SABSSM) 2017² and Demographic and Health Survey (DHS) 2016³. Since the survey sample size in each district is relatively small, routinely reported data about HIV prevalence among pregnant women attending their first ANC visit, extracted from the national District Health Management Information System (DHIS), are used to improve estimates of the spatial pattern of HIV.

ART coverage by district, age, and sex is estimated from household survey data about the presence of ARV biomarkers in HIV-positive survey respondents. Routinely reported ART coverage among pregnant women prior to first ANC visit is used as a covariate for the spatial pattern of ART coverage. The ART coverage and HIV prevalence are also calibrated so that total number on ART matches the number of adults and children accessing treatment in each district through public sector and private sector treatment provision, reported through the national DHIS and Council for Medical Schemes, respectively.

A challenge for estimating treatment coverage at district level is that persons may access ART services in a different district than their residence, for example if facilities are closer or perceived to provide better services. The model allows for a probability that resident PLHIV access ART in a neighbouring district. The prior assumption is that the large majority of PLHIV will access ART in their district of residence, but this probability can vary based on district data about the number receiving ART compared to HIV prevalence, ART coverage and population.

HIV incidence in 2017 is estimated from the proportion recently infected from the SABSSM 2017 survey. HIV incidence is determined by the district-level HIV prevalence and ART coverage such that the estimated HIV incidence is related to the prevalence of unsuppressed HIV in the population. The sex and age distribution of new infections is estimated using incidence rate ratios from Thembisa.

The next step of the model is to conduct a one-step projection of the population to March 2020. Population estimates are updated with Statistics South Africa district projections. The number of PLHIV is projected from 2017 to 2020 based on survival estimates by province, sex, and age group from Thembisa over the same period (which accounts for HIV disease progression and effects of ART scale up on reducing AIDS mortality). ART coverage is updated based on the number on ART in March 2020 from DHIS and medical scheme reporting. Finally, for programme planning purposes, a short term projection is calculated to March 2021 using the same methods. The number on ART is projected based on Thembisa model projection for rates of further scale up.

For paediatric estimates, where household survey data have low numbers of observed cases, the model uses the ratio of female adult HIV prevalence to child HIV prevalence from Thembisa provincial results to estimate the average district-level prevalence among children and the new paediatric infections between 2017 and 2020. Paediatric ART coverage is determined by the number of children receiving ART in each district and SABBSM survey data about ART coverage among children.

A preliminary version of the Naomi model was first developed in South Africa in 2019 and this report is the first release of Naomi model estimates. The model has been used in UNAIDS-supported national HIV estimates in several other sub-Saharan African countries. Future development of Naomi will extend the model for additional data sources and indicators including the HIV care cascade and prevention indicators.

Statistics South Africa. Mid-year population estimates 2019. 2019. Available: http://www.statssa.gov.za/publications/P0302/P03022019.pdf. Accessed 17 June 2020↩︎
Simbayi LC, Zuma K, Zungu N, Moyo S, Marinda E, Jooste S, et al. South African National HIV Prevalence, Incidence, Behaviour and Communication Survey, 2017. Cape Town: Human Sciences Research Council; 2019. Available: https://www.hsrcpress.ac.za/books/south-african-national-hiv-prevalence-incidence-behaviour-and-communication-survey-2017. Accessed 6 Nov 2019↩︎
Department of Health, Statistics South Africa, South African Medical Research Council, ICF. South Africa Demographic and Health Survey 2016. Pretoria; 2019. Available: https://www.dhsprogram.com/pubs/pdf/FR337/FR337.pdf. Accessed 19 March 2019↩︎